Probabilistic Descriptions of In-Situ Roof to Top Plate Connections in Light Frame Wood Structures

Failure of roof-to-wall top plate connections is a major cause of damage in light frame wooden structures subjected to high wind loads. Numerous post-storm investigations revealed that these connections fail below design wind loads [FEMA 2005; FEMA 2006; Kareem 1985]. This damage history highlights the need for understanding their resistance and behavior under such extreme loads. Earlier research studies replicated roof-to-wall top plate connections in the laboratory to study their ultimate capacity, but failed to capture the actual capacity of the connections in their "as built" in-situ condition. Furthermore, little effort has been made in the past to identify the probability distributions that describe the response and the ultimate capacity of these connections. This knowledge is needed to set the appropriate structural resistance criteria for the structural members and develop risk-consistent and reliable structural design methodologies for low-rise, wood-framed buildings. In this study, the uplift capacities of existing roof-to-wall toe nailed connections are determined and their associated probability distributions are presented. About 81 existing roof-to-wall top plate connections in light-framed wood residential structures built between 50 and 60 years ago were tested in accordance with ASTM D1761-2000 to determine their uplift performance. The results provided a robust statistical sample to assess the uplift capacity and probability distributions for typical toe-nailed roof-to-wall top plate connections in these residential structures. Results show that connections using 2-16d toenails have a mean ultimate capacity of 1.52 kN (342 lbs) and a coefficient of variation (COV) of 0.35. This COV value is found to be approximately twice that found from previous laboratory based studies lending motivation for conducting in-situ tests. Furthermore, cyclic testing of these connections also provided load-displacement relationships from which stiffness values were approximated. Results show that capacity and stiffness can plausibly be modeled as jointly lognormal.